Abstract
In the developing vertebrate retina, progenitor cell proliferation must be precisely regulated to ensure appropriate formation of the mature tissue. Cyclin kinase inhibitors have been implicated as important regulators of proliferation during development by blocking the activity of cyclin–cyclin-dependent kinase complexes. We have found that the p27Kip1 cyclin kinase inhibitor regulates progenitor cell proliferation throughout retinal histogenesis. p27Kip1 is upregulated during the late G2/early G1 phase of the cell cycle in retinal progenitor cells, where it interacts with the major retinal D-type cyclin–cyclin D1. Mice deficient for p27Kip1exhibited an increase in the proportion of mitotic cells throughout development as well as extensive apoptosis, particularly during the later stages of retinal histogenesis. Retroviral-mediated overexpression of p27Kip1 in mitotic retinal progenitor cells led to premature cell cycle exit yet had no dramatic effects on Müller glial or bipolar cell fate specification as seen with the Xenopus cyclin kinase inhibitor, p27Xic1. Consistent with the overexpression of p27Kip1, mice lacking one or both alleles of p27Kip1 maintained the same relative ratios of each major retinal cell type as their wild-type littermates. During the embryonic stages of development, when both p27Kip1and p57Kip2 are expressed in retinal progenitor cells, they were found in distinct populations, demonstrating directly that different retinal progenitor cells are heterogeneous with respect to their expression of cell cycle regulators.
The vertebrate retina is made up of six neuronal cell types and one glial cell type (for review, seeRodieck, 1998). In the murine retina, these diverse cell types are generated in a characteristic order during development (Young, 1985) from multipotent progenitor cells (Turner et al., 1990). It has been proposed that the birth order of retinal cell types reflects the unidirectional transition of progenitor cells through distinct stages of competence characterized by their ability to generate restricted subsets of retinal cell types (Cepko et al., 1996). Because of this developmental birth order, the proportion of cells that exit the cell cycle at each stage of development must be regulated carefully. If too many cells were to exit the cell cycle during the early stages of development, there might be an increase in the proportion of early-born cell types at the expense of later-born cell types.
In addition to the changes in progenitor competence over time during retinal histogenesis, there is evidence to suggest that at particular stages of development progenitors are a heterogeneous population that can exhibit biases in the fates adopted by their daughter cells (Alexiades and Cepko, 1997; Belliveau and Cepko, 1999; Belliveau et al., 2000). Therefore, along with the regulation of the total number of cells exiting the cell cycle over the course of retinal development, at any given stage of development the correct proportion of postmitotic daughter cells from each progenitor subpopulation also must be regulated carefully. If the newly postmitotic daughter cells were disproportionately derived from a subset of progenitor cells with a particular cell fate bias, then the proportion of cell types in the mature retina might be perturbed.
Previous research has demonstrated that the p57Kip2 cyclin kinase inhibitor is upregulated in progenitor cells during the late G1/G0 phase of the cell cycle and mediates cell cycle exit in the murine retina (Dyer and Cepko, 2000a). However, p57Kip2 is expressed in only a subset (∼16%) of mitotic progenitors between embryonic day (E) 14.5 and 17.5, raising the intriguing possibility that progenitor cells may use different mechanisms to exit the cell cycle during development. Although it has been reported that p27Kip1 is expressed in the embryonic retina (Zhang et al., 1998; Levine et al., 2000), a detailed analysis has not been performed on its role in the regulation of progenitor cell proliferation or cell fate specification. Moreover, it has not been established whether p27Kip1 plays a semi-redundant role with p57Kip2 in regulating progenitor cell proliferation or the two proteins function in distinct populations.
In the embryonic retina when both p27Kip1and p57Kip2 are expressed, we have found that they are expressed in distinct progenitor cell populations and are upregulated at different times in the cell cycle. Loss of one or both alleles of p27Kip1 was found to lead to extra rounds of cell division during development, but the distribution of the major cell types was not perturbed. In contrast to the p57Kip2-deficient mice, apoptosis did not occur when cells reentered the cell cycle but came much later at the end of retinal histogenesis. Retroviral overexpression of p27Kip1 in mitotic progenitor cells led to premature cell cycle exit, and as expected from premature exit, there was a reduction in the proportion of clones containing the cell types born at the end of retinal histogenesis: Müller glia and bipolar cells.
MATERIALS AND METHODS
Animals. C57BL/6, CD1, and ICR mice were purchased from Taconic Farms (Germantown, NY). p27Kip1 knock-out mice (Fero et al., 1996) were crossed to ICR or C57BL/6 mice with equivalent results. Genotypes were determined by performing PCR amplification of the wild-type and mutant alleles from tail DNA (Fero et al., 1996). Timed pregnant Sprague Dawley rats were purchased from Taconic Farms.
RNA isolation and RT-PCR assay. Three independent retinas were removed from staged embryonic (E14.5, E16.5, E18.5), postnatal (P0, P3, P6, P9, P12), and adult (6 weeks) ICR mice and immediately dissolved in 500 μl lysis solution (4 mguanidine thiocyanate, 25 mm sodium citrate, 0.5% Sarkosyl, 0.1 m β-mercaptoethanol). All three samples from each stage were analyzed, and a representative set is shown in Figure 1. RNA was prepared as described (Chomczynski and Sacchi, 1987). Expression of p27Kip1, cyclin D1, cyclin D3, and β-actin was analyzed in each sample by performing semiquantitative RT-PCR as described previously (Farrington et al., 1997). Sequence for the β-actin primers can be found in Farrington et al. (1997). Oligonucleotide primers for mouse p27Kip1 were (5′): 5′-AAACGTGAGAGTGTCTAACG-3′ (Tm = 51.4°C) and (3′): 5′-CCGTCTGAAACATTTTCTT-3′ (Tm = 51.2°C). Oligonucleotide primers for mouse cyclin D1 were (5′): 5′-ATGGAACACCAGCTCCTG-3′ (Tm = 55.3°C) and (3′):5′-CCAGACCAGCCTCTTCC-3′ (Tm = 54.5°C). Oligonucleotide primers for mouse cyclin D3 were (5′): 5′-TGTCCTGCAGAGTTTACTCC-3′ (Tm = 53.9°C) and (3′): 5′-GCAGGCAGTCCACTTCA-3′ (Tm = 54.9°C).
Immunohistochemistry, microscopy, and imaging. Retinal cryosections or dissociated cells (see below) were fixed in paraformaldehyde (4% in PBS), washed, and treated with hydrogen peroxide (1% in PBS) before incubation in blocking solution [PBS containing 0.1% Triton X-100 and 2% normal serum (Vector Laboratories, Burlingame, CA)]. For each of the antibodies listed below, the dilution used for retinal sections is listed first, followed by the dilution used for dissociated cell staining where applicable. Normal donkey serum was used for the following antibodies: anti-p27Kip1, clone 57 (mouse monoclonal, 1:50, 1:2000; Transduction Labs); anti-rhodopsin, Rho4D2 [mouse monoclonal, 1:250, 1:2000 (Molday and MacKenzie, 1983)]; anti-calretinin (mouse monoclonal, 1:500, 1:2000; Chemicon, Temecula, CA); anti-HNK-1, VC1.1 (mouse monoclonal, 1:1000, 1:5000; Sigma, St. Louis, MO); anti-syntaxin, HPC-1 (mouse monoclonal, 1:1000, 1:5000; Sigma); anti-calbindin-D28K, CL-300 (mouse monoclonal, 1:200, 1:2000; Sigma); anti-cyclin D1, 72–13G (mouse monoclonal, 1:500; Santa Cruz Biotechnology, Santa Cruz, CA); anti-FLAG, M2 (mouse monoclonal, 1:100; Sigma); and anti-bipolar antigen, 115A10 [mouse monoclonal, undiluted(Onoda and Fujita, 1987)] antibodies. Normal goat serum was used for the anti-cyclin D3 (rabbit polyclonal, 1:200; Santa Cruz Biotechnology); anti-choline aceltyltransferase (rabbit polyclonal, 1:400, 1:2000; Chemicon); anti-Chx10 [rabbit polyclonal, 1:1000, 1:5000 (C. Cepko, unpublished data)[; anti-cellular retinaldehyde binding protein (CRALBP) [rabbit polyclonal, 1:1000, 1:5000 (De Leeuw et al., 1990)]; and anti-cone opsins [rabbit polyclonal, 1:5000 (Wang et al., 1992; Chiu et al., 1994)] antibodies. Normal rabbit serum was used for the anti-p57Kip2, E-17 (goat polyclonal, 1:50, Santa Cruz Biotechnology) antibody. Biotin-conjugated secondary antibodies (donkey anti-mouse IgG, rabbit anti-goat IgG, goat anti-rabbit IgG; Vector Laboratories) were used at a dilution of 1:500 in blocking solution. After secondary antibody binding, an avidin–biotin–peroxidase complex (Vectastain ABC, Vector Laboratories) was incubated with the sections or dissociated cells followed by diaminobenzidine detection (Vector Laboratories), FITC tyramide, or Cy-3 tyramide detection (DuPont NEN, Wilmington, DE) according to the manufacturers' instructions (Bobrow et al., 1991). For some experiments, flurophor-conjugated tyramine compounds and reaction buffers were synthesized according to previous reports (Bobrow et al., 1991) with equivalent results. For nuclear staining, DAPI was added to the final wash solution at 0.0005%. Labeled cells were visualized using a Zeiss Axioplan-2 microscope with 10×, 20×, and 40× Plan Neofluar objectives or a 100× Plan Apochromat objective with adjustable iris. Images were captured with a Spot digital camera (Diagnostic Instruments). Confocal microscopy was performed using a Leica DM-RBE microscope equipped with a TCSNT true confocal scanner.
[3H] thymidine and BrdU labeling. To label retinal progenitor cells in S-phase, retinas were incubated in 1 ml explant culture medium containing [3H] thymidine [DuPont NEN; 5 μCi/ml (89 Ci/mmol)] or 10 μm bromodeoxyuridine (BrdU) (Boehringer Mannheim, Indianapolis, IN) for 1 hr at 37°C. Autoradiography and BrdU detection were performed as described previously (Morrow et al., 1998).
Retinal explant culture and dissociation. The procedure for explant culturing of mouse retinas has been described in detail previously (Dyer and Cepko, 2000a). Extensive characterization has demonstrated that retinal proliferation and differentiation are normal using this explant culture system (Dyer and Cepko, 2000a). Tissue dissociation was performed as described previously (Morrow et al., 1998).
Replication incompetent retroviral vector constructs and viral production. Oligonucleotides encoding the FLAG–His cassette were synthesized [for sequence, see Dyer and Cepko (2000a)], annealed, and cloned into the pNIN replication incompetent retroviral vector (Cepko, unpublished data) to make pNIN-E, the pLIA replication incompetent retroviral vector (Cepko et al., 1998) to make pLIA-E, or the pGFP vector to make pGFP-E. Mouse p27Kip1 was PCR amplified, sequenced, and cloned into pNIN-E, pLIA-E, and pGFP-E to generate pNIN-Ep27, pLIA-Ep27, and pGFP-Ep27, respectively. Oligonucleotide primers for p27Kip1 PCR amplification were as follows: p27-amino, 5′-TAGAGCGGCCGCATCTAACGTGAGAGTGTCT-3′ and p27-carboxy, 5′-TAGAGCGGCCGCCGTCTGGCGTCGAAGGCC-3′.
Xenopus p27Xic1 was PCR amplified, sequenced, and cloned into pLIA-E to generate pLIA-EXic1. Oligonucleotide primers for p27Xic1 were as follows: Xic1-amino, 5′-TAGAGCGGCCGCAGCTGCTTTDCCACATCGCC-3′ and Xic1-carboxy, 5′-TAGAGCGGCCGCTCGAATCTTTTTCCTGGG-3′.
To prepare high-titer retroviral stocks, the plasmid constructs were transiently transfected into a 293T ecotropic producer cell line (Phoenix-E) by calcium phosphate coprecipitation as described (Cepko et al., 1998). Supernatant containing the viral particles was harvested at 48 hr after transfection, and viral titer was determined on NIH-3T3 cells (Cepko et al., 1998). In vivo lineage analysis was performed as described previously (Turner and Cepko, 1987; Fields-Berry et al., 1992).
Recombinant p27Kip1 purification, coimmunoprecipitation, and immunoblotting. Recombinant histidine-tagged p27Kip1 was prepared using a baculovirus expression vector system (PharMingen, San Diego, CA) and purified on Ni2+-NTA agarose resin (Qiagen, Hilden, Germany) according to the manufacturer's instructions for nondenaturing conditions (Dyer and Cepko, 2000a). Recombinant proteins were used as positive controls for immunoprecipitation and immunoblotting experiments. For cyclin D1 coimmunoprecipitation, 10 P0 retinas from CD1 mouse pups were sonicated briefly in 2 ml 1× RIPA buffer (1× PBS, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 1 mm PMSF) containing a cocktail of protease inhibitors (Sigma), and phosphatase inhibitors (1 mm Levamisole, 2 mmNa2VO3, 1 mm NaF). This crude retinal lysate was cleared by spinning at 14,000 × g and protein-G Agarose preclearing was performed according to the manufacture's instructions (Santa Cruz Biotechnology). Anti-cyclin D1, C-20 (rabbit polyclonal, 1 μg; Santa Cruz Biotechnology) antibody was incubated with gentle inversion for 1 hr followed by a 1–2 hr incubation with protein-G Agarose. Washes and elution were performed according to the manufacturer's instructions (Santa Cruz Biotechnology). Crude retinal lysates, washes, and immunoprecipitates were separated on a 12% polyacrylamide gel containing SDS and transferred to nitrocellulose. Blocking, washing, and primary antibody incubations (anti-p27Kip1, 1:1000) were performed according to the manufacturer's instructions (Transduction Labs). The secondary biotinylated antibody (donkey anti-mouse IgG; Vector Laboratories) was used at a dilution of 1:2000. Amplification was achieved by incubating the immunoblot with an avidin–biotin–alkaline phosphatase complex (Vectastain-AP, Vector Laboratories) followed by nitro blue tetrazolium–5-bromo-4-chloro-3-indolyl phosphate detection (Vector Laboratories).
Apoptosis analysis. The colorimetric apoptosis detection system [terminal deoxynucleotidyl transferase-mediated biotinylated dUTP nick end labeling (TUNEL)] was used on 20 μm cryosections according to the manufacturer's instructions (Promega, Madison, WI).
Dissociated cell scoring and statistical methods. To evaluate the significance of differences in the proportion of cell types between wild-type, p27Kip1-heterozygous, and p27Kip1-deficient retinas, the mean and SD were calculated for counts of retinas from each genotype, and at test was performed. All p values are one-sided unless indicated otherwise.
RESULTS
p27Kip1 expression during development
As a first step toward understanding the kinetics of p27Kip1 mRNA expression over the course of retinal histogenesis, semiquantitative RT-PCR analysis was performed on three independent retinas from eight stages of development. Using primers specific for the p27Kip1coding sequence, mRNA was detected at E14.5 and persisted throughout development, peaking around P0 when the number of mitotic cells producing postmitotic daughter cells is the highest in the rodent retina (Fig. 1A) (Alexiades and Cepko, 1996). Notably, p27Kip1 expression was also found in the adult retina where there are no mitotic cells present (Fig.1A). For comparison, oligonucleotide primers specific for transcripts from the cyclin D1 and D3 genes were included in this analysis. It has been well established that cyclin D1 is the major D-type cyclin found in mitotic retinal progenitor cells during development (Fantl et al., 1995; Sicinski et al., 1995), and cyclin D3 is expressed in Müller glial cells of adult retina (Dyer and Cepko, 2000b; C. Ma and C. Cepko, unpublished observations). Because the percentage of mitotic cells decreased during development (Alexiades and Cepko, 1996), cyclin D1 mRNA expression tapered off such that in the mature retina, very little cyclin D1 was detected (Fig.1A). In contrast to cyclin D1, cyclin D3 was expressed at very low levels in the developing retina but was sharply upregulated during the late perinatal stages (Fig.1A).
The cellular distribution of the p27Kip1protein was examined by performing immunohistochemical staining on mouse retinas from six stages of development (Fig.1C–F) (data not shown). An antibody specific for cyclin D1 was included to label the mitotic retinal progenitor cells (Fig. 1B,G) (data not shown). Cells move within the developing retina according to cell cycle phase; mitosis occurs adjacent to the pigmented epithelium (PE), and S-phase occurs closer to the vitreal surface near the boundary between the inner neuroblastic layer (inbl) and the outer neuroblastic layer (onbl) (Sauer, 1937). The two gap phases (G1 and G2) mark the movement of cells between the PE and the inbl/onbl boundary. At E14.5, two populations of p27Kip1-expressing cells were detected (Fig. 1C). A subset of cells along the outer edge of the retina where newly postmitotic neurons fated to be cones and rods are found, as well as those on the inner surface where ganglion cells are differentiating, expressed high levels of p27Kip1 (Fig. 1C). A second group of p27Kip1-immunoreactive cells was detected throughout the onbl in the region where cyclin D1 is normally expressed (Fig. 1, compare B, C).
At later stages of development (Fig.1D–F), p27Kip1 was also expressed in gap phase cells along with newly postmitotic cells in the developing inner nuclear layer (INL) and ganglion cell layer (GCL). Cyclin D1 was expressed primarily in mitotic cells and appeared to be downregulated quickly in these newly postmitotic daughter cells (Fig. 1G) (data not shown). In the adult retina, p27Kip1 expression colocalized with cyclin D3 in Müller glial cells (Dyer and Cepko, 2000b).
Previous work has demonstrated that a subset of progenitor cells exiting the cell cycle in the embryonic retina upregulate p57Kip2 (Dyer and Cepko, 2000a). To determine whether p27Kip1 and p57Kip2 are found in distinct progenitor cell populations, double-label immunocytochemical staining was performed on E14.5 retinal sections with antibodies directed against p27Kip1 and p57Kip2. We found that these two proteins did not colocalize in the E14.5 retina (Fig.1H–J). Similarly, in the adult retina (Fig. 1K–M), these two proteins defined distinct, highly restricted populations of retinal neurons and glia (Dyer and Cepko, 2000b; Dyer and Cepko, 2001).
Timing of p27Kip1 upregulation during the cell cycle
To determine whether the onset of p27Kip1 expression within the cell cycle indicates that it might regulate progenitor cell proliferation, the expression of p27Kip1 was examined during different phases of the cell cycle. S-phase cells were pulse-labeled by incubating retinas with [3H]thymidine for 1 hr. After [3H]thymidine labeling, retinas were cultured as explants for various lengths of time, dissociated, and reacted with antisera specific for p27Kip1. After autoradiography, the proportion of [3H]thymidine-labeled cells expressing p27Kip1 was scored (Fig.2A–C, Table1). This analysis was performed at four stages of development (E14.5, E17.5, P0, and P2), spanning the period when p27Kip1 is expressed in retinal progenitor cells (Fig. 2D–G, Table 1). It was critical to examine all of these stages because the proliferation properties of retinal progenitor cells change during development in rodents (Alexiades and Cepko, 1996), and we wanted to determine whether the onset of p27Kip1expression during the cell cycle reflected those changes.
At E14.5, immediately after labeling (t = 0), none of the cells expressing p27Kip1 (0/270, 0%) were labeled with [3H]thymidine and therefore were not in S-phase (Fig. 2D, Table 1). Four hours later (t = 4), when many of the [3H]thymidine-labeled cells would have entered G2 (Alexiades and Cepko, 1996), some (27/328, 8.2%) [3H]thymidine-labeled cells expressed p27Kip1 (Fig.2D, Table 1). Progenitors in S-phase at the time of labeling should begin to enter G1 by 8 hr (t = 8) after labeling (Alexiades and Cepko, 1996). At that time point, a significant increase (57/361, 15.8%) in the proportion of [3H]thymidine-labeled cells expressing p27Kip1 was observed (Fig. 2D, Table 1). Later time points showed a slight increase in the proportion of double-labeled cells (Fig.2D, Table 1); however, the vast majority of retinal progenitor cells upregulated p27Kip1during the late portion of G2 or early part of G1, consistent with the p27Kip1 expression pattern seen at E14.5 (Fig. 1). Transcription is silenced during M phase so it is unlikely that p27Kip1 is upregulated during this phase of the cell cycle (Sanchez and Dynlacht, 1996).
From the earliest stages of retinal development in rodents when the first postmitotic daughter cells are being generated to the cessation of mitotic activity, the length of the cell cycle increases from ∼14 hr at E14.5 to ∼55 hr at P8 (Alexiades and Cepko, 1996). The timing of p27Kip1 upregulation in individual progenitor cells during the cell cycle may reflect this change in cell cycle kinetics, or cell cycle length may be intrinsically regulated. To distinguish between these two possibilities, a similar [3H]thymidine-labeling experiment was performed at E17.5, P0, and P2 (Fig.2E–G, Table 1). By E17.5 the timing of the onset of p27Kip1 expression was delayed by ∼4 hr as compared with the E14.5 labeling experiment (Fig.2E, Table 1), and in postnatal retinal progenitor cells (P0, P2) there was an even longer delay (14 hr) in the accumulation of [3H]thymidine-labeled cells expressing p27Kip1 (Fig.2F,G, Table 1).
Cyclin D1 is the major D-type cyclin found in mitotic retinal progenitor cells of the murine retina and is required for progenitor cell proliferation (Fantl et al., 1995; Sicinski et al., 1995; Ma et al., 1998). Because of this central role in regulating retinal progenitor cell proliferation, a coimmunoprecipitation experiment was performed to determine whether p27Kip1interacts with cyclin D1 in vivo. Protein lysates from P0 retinas were incubated with an anti-cyclin D1 antibody, immunoprecipitated, separated by SDS-PAGE, and immunoblotted with an antibody specific for p27Kip1. Cyclin D1 and p27Kip1 formed a complex in lysates from retinas when the number of mitotic cells was highest (Alexiades and Cepko, 1996) (Fig. 2H).
Retroviral-mediated overexpression of p27Kip1 in mitotic retinal progenitor cells
To test whether p27Kip1 expression is sufficient to drive retinal progenitor cells out of the cell cycle, and to examine any effects of p27Kip1overexpression on cell fate specification, three replication incompetent retroviruses containing the p27Kip1 cDNA were generated (Fig.3A). One of these viral constructs (Fig. 3A,pNIN-E(Kip1 )) contains a nuclear β-galactosidase reporter gene and is ideally suited for analyzing the effects of p27Kip1 overexpression on progenitor cell proliferation (Dyer and Cepko, 2000a). The second viral construct (Fig.3A, pLIA-E(Kip1) ) contains an alkaline phosphatase reporter gene and is similar to constructs used previously for in vivo lineage analysis in the rodent retina (Cepko et al., 1998). A retroviral construct encoding green fluorescent protein (GFP) was also generated for coimmunolocalization experiments. By taking advantage of the epitope tag (FLAG) encoded on the amino terminus of p27Kip1 in these vectors (Fig.3A), we demonstrated that significant levels of p27Kip1 protein were expressed from pNIN-EKip1 and pLIA-EKip1 (Fig. 3B). Furthermore, infected fibroblasts (NIH-3T3) exited the cell cycle but did not undergo apoptosis (data not shown). Finally, immunolocalization of the FLAG epitope in 293T cells transfected with a similar retroviral construct encoding GFP (Fig. 3A,pGFP-E(Kip1) ) demonstrated that most cells (189/200, 94%) expressing GFP also express nuclear-localized p27Kip1 (Fig.3C).
To examine the effects of p27Kip1overexpression on progenitor cell proliferation, E14.5 murine retinas (n = 43) were infected with NIN-EKip1or NIN-E and cultured for 10 d as explants (see Materials and Methods). After this culture period, retinas were stained for β-galactosidase expression and sectioned, and the size of clones derived from single infected progenitor cells was scored (Fig.3D–F). A distribution in clone size ranging from 1 to 29 cells was observed in retinas infected with the control virus (NIN-E) and from 1 to 16 cells for the retinas infected with NIN-EKip1 (Fig.3F). The proportion of single cell clones in retinas infected with NIN-EKip1 (51/102, 50 ± 2.7%) was significantly increased compared with those infected with the control virus (50/160, 30 ± 2.2%) (p< 0.01) (Fig. 3F). Furthermore, the proportion of large clones (more than five cells) was significantly higher in retinas infected with NIN-E (50/160, 33 ± 2.4%) than NIN-EKip1 (13/102, 13 ± 0.3%) (p < 0.003) (Fig. 3F).
Although these data suggest that overexpression of p27Kip1 may be sufficient to drive retinal progenitor cells out of the cell cycle, it is possible that the smaller clone size resulted from apoptosis as a consequence of p27Kip1 overexpression. Therefore we compared the kinetics of clone size distribution in retinas infected with NIN-E with those infected with NIN-EKip1 over the course of several days in culture (Fig. 3G). If apoptosis played a significant role in the reduction of the size of clones derived from progenitor cells infected with NIN-EKip1, there might be an early peak in clone size followed by a decrease attributable to apoptosis. Alternately, if p27Kip1overexpression simply forced progenitor cells out of the cell cycle, then the decrease in clone size should remain relatively constant over the culture period. Data from >600 clones suggest that large clones are not generated and then pruned by apoptosis to reduce the clone size resulting from progenitor cells overexpressing p27Kip1 (Fig. 3G). As additional support for this conclusion, no increase in the proportion of apoptotic nuclei was detected in clones (n > 50) expressing p27Kip1 using the TUNEL assay (data not shown).
In the Xenopus retina, transfection of progenitor cells with a plasmid encoding p27Xic1 led to an increase in the number of Müller glial cells at the expense of bipolar neurons (Ohnuma et al., 1999). To test whether p27Xic1 could induce a similar alteration in cell fate specification in the rodent retina, a high titer stock of the LIA-EXic1 retrovirus was injected into the left eye of newborn rat pups; the control LIA-E virus was injected into the contralateral eye (Cepko et al., 1998). Retinas were harvested after complete retinal development (P21), stained for alkaline phosphatase expression, and sectioned. Clones of cells derived from individually infected retinal progenitor cells (Fig.4A–F) were scored for clone size and clone composition (Fig.4G,H) (Turner and Cepko, 1987;Fields-Berry et al., 1992). Similar to the data from Xenopus(Ohnuma et al., 1999), the proportion of clones containing bipolar interneurons was decreased from 19% (32/165) for LIA-E to 8% (35/416) for LIA-EXic1 (Fig. 4G). If all of the infected cells are treated as a population, then the proportion of bipolar cells was decreased from 9.4% (32/337) for LIA-E to 6.2% (35/563) for LIA-EXic1. However, the proportion of clones containing Müller glia was only slightly increased from 8% (14/165) for LIA-E to 10% (43/416) for LIA-EXic1 (Fig. 4G). This difference is more pronounced when the proportion of cells is compared [4.1% (14/337) for LIA-E to 7.6% (43/563) for LIA-EXic1] rather than the proportion of clones containing those cells. Furthermore, mitotic retinal progenitor cells prematurely exited the cell cycle as indicated by an increase in the proportion of single rod clones among the clones that contain only rods (47% for LIA-E and 81% for LIA-EXic1).
To determine whether overexpression of the mouse p27Kip1 was sufficient to force retinal progenitor cells out of the cell cycle in vivo and whether Müller glial/bipolar cell fate specification was perturbed, we performed a similar lineage study with LIA-EKip1. The titer of the LIA-EKip1 retrovirus was significantly lower on 3T3 cells (∼1 × 10−6/ml) than that obtained for LIA-EXic1(∼5 × 10−6/ml). This disparity in titer was also reflected in the average number of clones per retina (∼12 clones per retina for LIA-EKip1 and ∼50 clones per retina for LIA-EXic1) for the in vivolineage analysis. Although we could not detect any cytotoxicity/apoptosis as a result of p27Kip1 misexpression (Fig. 3, and see above), it is possible that a subset of cells was selectively killed as a result of persistent p27Kip1 expression. Low titer notwithstanding, there was an obvious reduction in the size of clones derived from progenitor cells infected with LIA-EKip1 (82% of rod-only clones were single-rod clones as compared with 63% for LIA-E). As expected (see Discussion), on the basis of this premature cell cycle exit, there was a slight reduction in the percentage of clones containing Müller glial cells as well as clones containing bipolar cells (Fig.4H). The proportion of Müller glial cells among all the infected cells was similarly reduced from 5.3% (24/446) for LIA-E to 2.7% (8/388) for LIA-EKip1. Bipolar cells were reduced from 8.7% (39/446) for LIA-E to 5.5% (16/388) for LIA-EKip1.
Cell cycle exit in the p27Kip1-deficient retinas
Mice carrying a targeted disruption of the p27Kip1 gene have been described previously and were found to exhibit multiple organ hyperplasia and increased body size as a result of increased proliferation (Fero et al., 1996; Kiyokawa et al., 1996; Nakayama et al., 1996). To test whether retinal progenitor cells undergo additional rounds of cell division in the absence of p27Kip1, a BrdU pulse-labeling experiment was performed (Fig.5A,B). At least 500 cells were scored from 8–12 retinas from five stages of development and in adult retinas (Fig. 5C) (data not shown). After scoring, genotypes were determined, and the data from the wild-type, p27Kip1 heterozygous, or p27Kip1-deficient animals were averaged (Fig. 5C). The proportion of mitotic cells observed at E14.5 in the p27Kip1-deficient retinas was significantly higher (40 ± 2.8%) than that of their wild-type littermates (26 ± 5.1%; p < 0.007) (Fig.5C). The proportion of mitotic cells in retinas from p27Kip1+/− animals (32 ± 1.7%) was intermediate between the data for the wild-type and knock-out mice (Fig. 5C). At E16.5, P0, P3, and P10, a similar pattern was observed (Fig. 5C). No mitotic cells were observed in retinas from adult animals at 3 weeks or 3 months of age for the p27Kip1-deficient or p27Kip1-heterozygous mice (data not shown).
Apoptosis in the retinas from p27−/− and p27+/− mice
As with the p27Kip1-deficient retinas, an increase in the proportion of mitotic cells was observed in the retinas from p57Kip2 knock-out mice (Dyer and Cepko, 2000a). This increased proliferation was accompanied by an increase in apoptosis during the stage when p57Kip2 was normally expressed and compensated for the extra cells in the p57Kip2 knock-out retina. To test whether a similar compensation mechanism was occurring in the p27Kip1-deficient or p27Kip1-heterozygous retinas, a TUNEL assay was performed on retinas from each stage of development examined above for BrdU labeling. During the early stages of development (E14.5 and E16.5), very few apoptotic nuclei were observed in any of the animals (data not shown). However, as development progressed the presence of an increased proportion of apoptotic nuclei was apparent in the p27Kip1−/−and p27Kip1+/− retinas as compared with their wild-type littermates. This difference was most significant at P10.5 (Fig. 5D–H). The proportion of apoptotic nuclei appeared to be greater in the p27Kip1+/− retinas than the p27Kip1−/−retinas, particularly in the outer nuclear layer (Fig. 5E).
Quantitation of the major retinal cell types in the p27+/− and p27−/−mice
To test whether the additional proliferation and apoptosis of the p27Kip1 mutant retina resulted in aberrant production or survival of particular cell types, the proportion of several classes of retinal cell types was examined in adult retinas from wild-type, p27Kip1+/− and p27Kip1−/−mice. Retinas from 6-week-old mice from a cross of p27+/−parents were dispersed, plated, and stained with various cell type-specific antibodies (Fig.6A–E). Mice were genotyped after cell counting, and data from each genotype were pooled to obtain the mean and SD for each group of samples (Fig.6F). Rod photoreceptors (Fig. 6A) constitute the majority of cells in the adult murine retina, and no significant difference in the proportion of rhodopsin-immunoreactive photoreceptors was found in mice lacking one or both alleles of p27Kip1 (Fig. 6F). The proportion of Müller glial cells, as measured by CRALBP immunoreactivity (Fig. 6E), was not decreased in the retinas from mice deficient for p27Kip1(Fig. 6F). We did find, however, a dramatic (10- to 20-fold) increase in the proportion of Müller glial cells expressing glial fibrillary acidic protein, which is an intermediate filament protein found in Müller cells undergoing reactive gliosis (Dyer and Cepko, 2000b). The other major retinal cell types, 115A10 and Chx10 immunoreactive bipolar cells (Fig.6D) (data not shown), syntaxin immunoreactive amacrine cells (Fig. 6C), and calbindin immunoreactive horizontal cells (Fig. 6B) were present in approximately the same proportion in the retinas from all of the mice examined (Fig. 6F). Furthermore, amacrine cell subpopulations (calretinin, calbindin, ChAT, parvalbumin, and p57Kip2) were unaffected in the retinas from mice lacking one or both alleles of p27Kip1 (data not shown).
Organization of cell types in the p27Kip1-deficient retinas
To determine whether the major retinal cell types were organized appropriately into the correct laminas of the retinas in the p27Kip1-deficient mice, immunohistochemical staining was performed on retinal sections using the same antibodies described above. We found that the boundaries between the cellular layers of the retina in the p27Kip1-deficient mice were disrupted. The cell bodies of the rhodopsin-immunoreactive photoreceptors were found outside the outer limiting membrane, and the photoreceptor outer segments were often missing in those regions (Fig.7A,B). The photoreceptor layer of the p27+/−retinas did not exhibit the type of retinal dysplasia (Nakayama et al., 1996) seen in the p27−/−retinas. However, the boundaries between the outer nuclear layer and the inner nuclear layer did not appear as regular as in the wild-type littermates (data not shown). Additionally, bipolar interneurons (Fig.7C,D) and horizontal cells (Fig.7E–H) were displaced from their normal positions in the INL. Amacrine cells (data not shown) and amacrine cell subpopulations [calbindin (Fig.7G,H); calretinin (Fig.7I,J); p57Kip2 (Fig.7K,L)] appeared to be localized to the correct region of the INL, yet their overall organization was not as regular as that seen in the wild-type retinas.
DISCUSSION
We have presented several lines of evidence to suggest that p27Kip1 is an important regulator of retinal progenitor cell proliferation during development. p27Kip1 was found to be upregulated during the late G2 or early G1phase of the cell cycle, overexpression of p27Kip1 in mitotic retinal progenitor cells led to premature cell cycle exit, and an increase in the proportion of mitotic cells was observed in the retinas from mice lacking one or both alleles of p27Kip1. Surprisingly, the proportion of the major retinal cell types in the mature retinas from p27Kip1-deficient mice was normal, suggesting that there was compensation for the extra rounds of cell division. In fact, more apoptosis was found in the p27Kip1−/−retinas, most likely accounting, at least in part, for this compensation. Overexpression of p27Kip1 using transduction via a retrovirus vector in vivo and in vitro led to smaller clones, suggesting that p27Kip1 is not only required for proper exit from the cell cycle but is sufficient to induce it. However, in contrast to the Xenopusp27Xic1 cyclin kinase inhibitor, mouse p27Kip1 did not lead to any obvious perturbation in cell fate determination that could not be explained by the premature cell cycle exit of retinal progenitor cells. Significantly, p27Kip1 and p57Kip2, two regulators of cell cycle exit of the Cip/Kip family, were expressed in distinct retinal progenitor cell populations and upregulated at different times in the cell cycle.
Retinal progenitor cells use at least two different cyclin kinase inhibitors to exit the cell cycle
Previous work has demonstrated that the p57Kip2 cyclin kinase inhibitor mediates cell cycle exit in a restricted subset (∼16%) of embryonic retinal progenitor cells (Dyer and Cepko, 2000a). We have shown here that the p27Kip1 cyclin kinase inhibitor is expressed in a distinct population of retinal progenitor cells during embryonic development. Not only were these two proteins expressed in different groups of progenitor cells, they were upregulated during different phases of the cell cycle. p27Kip1 expression was detected within 8 hr of S-phase at E14.5, which is consistent with the late G2 or early G1 phase of the cell cycle. Because the length of the cell cycle increased during development (Alexiades and Cepko, 1996), the timing of p27Kip1 upregulation after S-phase was similarly delayed. This may indicate that upregulation of p27Kip1 in retinal progenitor cells occurs at the same phase of the cell cycle regardless of cell cycle length. In contrast to the timing of p27Kip1upregulation, p57Kip2 expression was not detected until 16 hr after S-phase, which is consistent with expression in the late G1 or G0 phase of the cell cycle (Dyer and Cepko, 2000a).This is the first example of retinal progenitor heterogeneity with respect to the mechanism of cell cycle exit. For example, progenitor cells may have the ability to produce different daughter cell types because of the usage of different cyclin kinase inhibitors (p27Kip1 vs p57Kip2) (Alexiades and Cepko, 1997; Dyer and Cepko, 2000a). Alternatively, work on Dictyostelium has demonstrated that cells respond differently to the same stimuli depending on cell cycle phase (Gomer and Ammann, 1996). Thus, the time within the cell cycle that a retinal progenitor cell decides to produce a postmitotic daughter cell, or to become postmitotic, may influence which cyclin kinase inhibitor is upregulated. For example, if a progenitor cell decides to produce a postmitotic daughter between the late G2 and early G1 phases of the cell cycle, then p27Kip1 might be upregulated, whereas if the decision to exit the cell cycle is delayed by several hours (late G1 phase), a progenitor cell might upregulate p57Kip2. Because changes in the competence of retinal progenitor cells to produce different retinal cell types occurs during retinal development concomitant with changes in the kinetics of cell cycle and mitotic fate of daughter cells, it is possible that all of these changes are linked.
In retinas from p27Kip1-deficient mice, the proportion of mitotic cells was increased in comparison to their wild-type littermates. However, this difference was somewhat lower than expected considering the broad expression of p27Kip1 during development. Thus, in the absence of p27Kip1, retinal progenitor cells may use an alternative, semi-redundant mechanism to exit the cell cycle. The most obvious possibility would be the presence of one or more additional cyclin kinase inhibitors. Consistent with this model, we have found that in addition to p27Kip1and p57Kip2, there are two other cyclin kinase inhibitors expressed in the developing mouse retina (our unpublished observations). It is also possible that in the absence of p27Kip1, proteins that do not normally act as cyclin kinase inhibitors may serve that role. Specifically, work on mouse embryonic fibroblasts lacking p27Kip1 demonstrated that a member of the retinoblastoma (Rb) family of proteins can serve as a cyclin kinase inhibitor in those cells (Zhu et al., 1995; Woo et al., 1997; Coats et al., 1999). All three Rb family members are expressed in the murine retina (our unpublished observations), and one or more of these molecules may mediate cell cycle exit in the absence of p27Kip1. Finally, our expression studies revealed that cyclin D1 is rapidly downregulated in newly postmitotic daughter cells. Therefore, the precise timing of cell cycle exit may be a two-step process: downregulation of cyclin D1 and upregulation of a cyclin kinase inhibitor. In the absence of p27Kip1, progenitor cells may still eventually exit the cell cycle simply through their normal process of downregulating cyclin D1.
Extra cells in the p27Kip1-deficient retinas are eliminated by apoptosis during the late perinatal stages of development
Because of the birth order of retinal cell types during development, perturbations in progenitor cell proliferation could affect the proportion of one or more of these cell types in the mature tissue. The increase in mitoses observed throughout development in the p27Kip1-deficient retina could lead to a large cumulative change in retinal cell number such that a change in the proportions of retinal cell types might have been observed in the adult. Surprisingly, there was no change in the proportions of retinal neurons or glia in the adult retina. Previous work on the p57Kip2-deficient retina demonstrated that inappropriate S-phase entry was quickly followed by apoptosis during the embryonic period when p57Kip2 is expressed (Dyer and Cepko, 2000a). However, very little apoptosis was detected throughout much of development in the p27Kip1-deficient retina, although there was a greater than normal proportion of cells in S-phase. In contrast to our observations of the timing of apoptosis in the p57Kip2 deficient retinas, an enormous number of apoptotic nuclei were observed in the retinas from mice lacking one or both alleles of p27Kip1 after proliferation was complete (P10.5). This may indicate that the extra cells generated during retinal development were not eliminated when they reentered the cell cycle, as seen in the p57Kip2-deficient retina, but were eliminated all at once postnatally.
This difference in the timing of apoptosis in p27Kip1- and p57Kip2-deficient retinas may indicate that the two genes play different roles. p57Kip2 may be required to prevent reentry of cells into S-phase after they have entered G0. The observation that p57Kip2-deficient cells undergo apoptosis after they have migrated to the inner retina (Dyer and Cepko, 2000a) where they most likely would be beginning to differentiate as amacrine cells supports the notion that they are attempting to enter S-phase from G0. This type of behavior apparently leads to immediate apoptosis, not only in the p57Kip2-deficient retinas, but also in the CNS of Rb-deficient mice (Lee et al., 1992). The role played by p57Kip2 thus seems distinct from that of p27Kip1 in terms of two criteria: (1) kinetics of synthesis during the cell cycle (late G1/G0 for p57Kip2 and late G2/early G1 for p27Kip1 and (2) timing of apoptosis (immediate for p57Kip2-deficient retinas and delayed for p27Kip1-retinas). The delay in apoptosis for p27Kip1-deficient retinas suggests that these cells simply fail to exit the cell cycle and continue to proliferate somewhat normally, as opposed to reentering the cell cycle from an inappropriate stage of the cell cycle or using an aberrant mechanism. The extra cells that are generated in the p27Kip1-deficient retinas are therefore “normal” but in excess. The excess is then partially or completely eliminated at the end of development.
Despite the normal proportion of retinal cell types in mice lacking one or both alleles of p27Kip1, we found that the organization of these cell types was perturbed. These defects occurred in regions of retinal dysplasia described previously (Nakayama et al., 1996). We have since found that retinal dysplasia results from reactive gliosis involving Müller glial cells during development (Dyer and Cepko, 2000b). That is, disruptions in the outer limiting membrane, which is made up of Müller cell apical microvilli, probably result in the retinal disorganization described here. Furthermore, vascular defects seen in the retinas from p27Kip1-deficient mice (Dyer and Cepko, 2000b) may also be a contributing factor.
p27Kip1 does not play a direct role in cell fate specification or differentiation in the murine retina
Significant evidence is accumulating that cyclin kinase inhibitors can influence developmental processes beyond their prescribed role in proliferation control (Zhang et al., 1997; Ohnuma et al., 1999; Dyer and Cepko, 2000a). In the Xenopus retina, overexpression of p27Xic1 led to an increase in the proportion of Müller glial cells and a reduction in the proportion of bipolar cells (Ohnuma et al., 1999). When p27Xic1 expression was blocked, a decrease in Müller glial cells was observed along with an increase in bipolar interneurons. We found that overexpression of p27Xic1 in murine retinal progenitor cellsin vivo led to a reduction in clone size and a decrease in the proportion of clones containing bipolar cells and a modest increase in the proportion of clones containing Müller glia. However, when the total population of cells infected was considered rather than the clonal composition, there was a decrease in bipolar cells from 9.4% (32/337) for LIA-E to 6.2% (35/563) for LIA-EXic1 and an increase in the Müller glial cells from 4.1% (14/337) for LIA-E to 7.6% (43/563) for LIA-EXic1. These approximately twofold differences are similar to the data obtained inXenopus for the total population of transduced cells at a similar stage of development (stage 21–24) (Ohnuma et al., 1999). Therefore, our data indicate that p27Xic1has a similar affect on rodent retinal progenitor cell proliferation and specification/differentiation as was shown previously forXenopus retinal progenitor cells (Ohnuma et al., 1999).
The murine cyclin kinase inhibitor, p27Kip1, also led to a reduction in clone size and a reduction in the proportion of clones containing bipolar cells. However, in contrast to the Xenopus Xic1 protein, mouse p27Kip1 did not increase the proportion of clones containing Müller glial cells but actually decreased them slightly. Considering that Müller glial cells and bipolar cells are among the last cell types to be generated during retinal histogenesis, premature cell cycle exit should reduce the proportion of clones containing those cell types. Furthermore, the peak period of rod photoreceptor genesis occurs just before the peaks for bipolar and Müller glial cells. Thus, it is not surprising that premature cell cycle exit mediated by p27Kip1 would lead to an increase in the proportion of clones containing rod photoreceptors, as we observed. Indeed, misexpression of other cyclin kinase inhibitors in the developing rodent retina has also led to a reduction in bipolar and Müller glial cells accompanied by an increase in the proportion of clones containing rod photoreceptors (our unpublished observations). These data, combined with the aforementioned observation that the knock-out retinas had no major defect in the proportion of any of the retinal cell types, suggest that p27Kip1is not likely to play a direct role in cell fate specification in the murine retina.
There are several possible models for the role of p27 in retinal development that can explain the differences between Xic1 and Kip1 after misexpression in the Xenopus and murine retina. At the moment, it is not clear which model is correct. When overexpressed in the rodent retina, p27Kip1 and p27Xic1 gave different results regarding the number of Müller glia, suggesting that the two proteins are different with respect to their ability to induce this cell type. In contrast, in Xenopus, both proteins increased the number of Müller glia, which would suggest that they are similar in their ability to induce this cell type. When one examines the loss of function data, it is supportive of the overexpression data in each organism. Loss of function in the murine retina had no obvious effect on cell fate specification, whereas a reduction in the expression of p27Xic1 in the Xenopus retina did affect cell fate specification. Because Xic1 does not appear to have an ortholog in mammals and shares distinct sequence homology regions with different members of the mammalian Cip/Kip family (Ohnuma et al. 1999), it could be that the two proteins play different roles in their respective organisms. This may part be explained in part by the rapidity with which the Xenopus retina is built, relative to the murine retina. In Xenopus, nearly half of the total number of cells, representing all of the major cell types, are produced in the amount of time it takes to progress through one round of cell division in mice (∼10 hr). In mice, retinal histogenesis takes well over 2 weeks (Young, 1985), which is equivalent to ∼8–10 rounds of cell division (Alexiades and Cepko, 1996). Thus, the regulation of proliferation, the consequences of altered proliferation, and any changes in progenitor cell competence to make different cell types might be different in the different organisms.
Cyclin kinase inhibitors play multiple, distinct roles in the formation and maintenance of a healthy retina
We have recently learned a great deal about the roles that cyclin kinase inhibitors can play in the vertebrate retina. It was not surprising to find, as we have shown here, that cyclin kinase inhibitors regulate progenitor cell proliferation during retinal development (Ohnuma et al., 1999; Dyer and Cepko, 2000a). However, the evidence for progenitor cell heterogeneity in terms of cell cycle exit (p27Kip1 vs p57Kip2 and possibly cyclin D1 vs cyclin D3) was unexpected and important for our understanding of retinal development. Beyond proliferation control, cyclin kinase inhibitors can also regulate cell fate specification and differentiation in the retina (Ohnuma et al., 1999; Dyer and Cepko, 2000a). In addition to these developmental processes, recent findings have shown that a cyclin kinase inhibitor (p27Kip1) is important for the initial response to injury in the adult retina (Dyer and Cepko, 2000b). Significantly, downregulation of p27Kip1 is the earliest molecular event identified to date characteristic of Müller glial cells undergoing reactive gliosis. When taken together, these related studies and the data presented here indicate that cyclin kinase inhibitors can play diverse and often unexpected roles in the developing and mature vertebrate retina.
Footnotes
M.A.D. was supported by National Research Service Award fellowship EY06803-02 and the Charles H. Revson Foundation Fellowship for Biomedical Research. This work was supported by National Institutes of Health Grant EY0-8064. We thank Dr. M. H. Baron for many helpful discussions and support throughout this project; Drs. S. Elledge, W. Harper, P. Zhang, and W. Harris for cDNAs; Dr. L. H. Tsai for knock-out mice; M. Peters for critical reading of this manuscript; and J. Zitz, M. Peters, and L. Rose for technical support.
Correspondence should be addressed to Constance L. Cepko, Department of Genetics and Howard Hughes Medical Institute, Harvard Medical School, 200 Longwood Avenue, Boston, Massachusetts 02115. E-mail: cepko{at}rascal.med.harvard.edu.